The Theory of Planned Behavior (TPB) explains human behavior, suggesting that intention shaped by attitudes, norms, and perceived control predicts online purchasing activities, among other influencing factors (Pedrinelli et al., 2024). The Theory of Planned Behavior (TPB) has been widely applied to understand and predict various behaviors across different domains. In the context of online food shopping, additional factors such as food-specific features (e.g., quality, freshness, delivery time, and packaging) may further impact consumer selections. While the TPB has been widely employed to analyze and predict many behaviors across multiple domains, its application to food purchasing behavior gives distinct insights (Martini et al., 2023; Rozenkowska, 2023). Food-related attributes present specific consumer concerns, such as the demand for freshness and convenience, which differentiate food transactions from other sorts of online buying. By integrating these food-specific characteristics into the TPB framework, a more comprehensive knowledge of online food purchasing behavior may be gained, distinguishing it apart from studies focusing on other types of consumer behavior (Dang-Van and Nguyen, 2025; Liao et al., 2025; Martini et al., 2023).

Attitude, defined as an individual’s evaluation of a behavior, significantly shapes behavioral intentions across various domains (Pedrinelli et al., 2024), including online food purchasing. Positive attitudes toward a service increase the likelihood of purchase intentions (Chen et al., 2020; Kim and Kim, 2022). Building on this theoretical background, the following hypotheses are proposed to explore the impact of key factors on online food purchase intention.

Subjective norm is the perceived social pressure to perform a behavior, shaped by the expectations of others, It along with attitude, Social approval and recommendations from peers influence individual purchase intentions, continued usage and repeat purchase (Lin et al., 2018; Pedrinelli et al., 2024), with positive societal perceptions encouraging the adoption of new business models in online food services (Fu et al., 2020; Pillai et al., 2022; Shanbhag et al., 2023), it significantly shapes consumer behavior in online food shopping. Based on this and the explanation in the introduction above, the following hypothesis can be formulated:

Perceived control refers to individuals’ perception of the ease or difficulty in performing a behavior, influenced by factors like quality and shelf life, and positively affects online purchase intentions by shaping risk judgment and interest (Ozkara et al., 2017; Pedrinelli et al., 2024), Perceived control, reflecting an individual’s confidence in their ability to engage in a behavior, is crucial for behavioral intentions. In online food shopping, consumers’ perception of control over usage processes influences their decision to use the service and recommend it to others (Ariño et al., 2011; Chen et al., 2020; Hagger et al., 2022). Based on this and the explanation in the introduction above, the following hypothesis can be formulated:

Convenience motivation in online shopping refers to the ease of use, saving time, and avoiding offline shopping challenges, which positively influences purchase intention, especially in online food purchasing through home delivery (Campo and Breugelmans, 2015; Alrawad et al., 2023; Muangmee et al., 2021; Guo et al., 2021), Convenience, defined by time value, ease, flexibility, and effort, is crucial in shaping consumer satisfaction and future intentions (Jalil et al., 2024; Srivastava and Thaichon, 2023), Based on this and the explanation in the introduction above, the following hypothesis can be formulated:

Service quality is crucial for customer satisfaction and organizational success, often assessed using the SERVQUAL model. High service quality boosts customer attitudes and purchase intentions, while poor quality has a negative impact. It is evaluated based on service, delivery process, and environment, with adaptations needed for specific contexts (Sohail and Hasan, 2021; Ariffin et al., 2021; Yang and Hu, 2022; Uzir et al., 2021). Service quality, including factors like reliability, responsiveness, and trust, significantly impacts consumer purchase intentions, fostering positive attitudes, loyalty, and repeat purchases, Maintaining high service standards is essential for enhancing consumer satisfaction and long-term loyalty (Bhati et al., 2022; Uzir et al., 2021; Maharsi et al., 2021; Bello et al., 2021; Jadil et al., 2022). Based on this and the explanation in the introduction above, the following hypothesis can be formulated:

Price saving orientation focuses on minimizing costs, with price reductions enhancing perceived value and influencing consumer decisions, particularly in online food delivery services (Nagle, 2024; Liu et al., 2023). The availability of online information allows price comparisons, with high perceived value and lower prices driving purchase decisions. Price-saving orientation and product quality influence customer satisfaction, choice, and purchase intentions (Allah Pitchay et al., 2022; Change, 2023; Giningroem et al., 2023; Troise et al., 2021). Based on this and the explanation in the introduction above, the following hypothesis can be formulated:

App quality refers to user assessments of factors like customization, interactivity, security, and convenience (Yang et al., 2022), High-quality apps, offering features like ease of use, flexible payment options, and effective support, enhance trust, satisfaction, and purchase intention. A well-designed app ensures reliable, up-to-date information, good reputation and a positive user experience (Qalati et al., 2021). Website and app quality significantly affect customer satisfaction and purchase intention, with key factors like system quality, information quality, and service quality driving these outcomes. App design, including esthetics and content quality, plays a crucial role in user experience and engagement. Lack of useful information, even in a well-designed app, can hinder consumer usage. The quality of app information is linked to trust, satisfaction, and purchase intention (Dung et al., 2024; Foroughi et al., 2024; Gai et al., 2024; Lin et al., 2024). Based on this and the explanation in the introduction above, the following hypothesis can be formulated:

Perceived risk refers to the uncertainty consumers face regarding potential negative outcomes of a purchase, with higher risks in online shopping compared to physical stores. It significantly influences consumer attitudes, behavior, trust, and the adoption of new products, particularly in the context of online food purchasing, which is considered more risky (Muharam et al., 2021; Essiz et al., 2023; Wang et al., 2022; Szymkowiak et al., 2021). Perceived risk, including financial, product, security, time, and psychological risks, significantly affects consumer attitudes and purchase intentions in online food shopping (Pillai et al., 2022; Rathi et al., 2024), Security risk is a primary barrier, while trust in sellers can mediate this relationship (Munikrishnan et al., 2023). Perceived benefits and risk-reducing strategies, such as brand reputation and certifications, can mitigate these concerns, enhancing purchase intentions (Nam et al., 2019). Based on this and the explanation in the introduction above, the following hypothesis can be formulated:

Intention to purchase indicates a consumer’s likelihood to buy, influenced by factors like attitude, behavior, and trust. It is shaped by product quality, service, security, and loyalty, and can be impulsive and unpredictable in E-commerce (Chung et al., 2021; Liu et al., 2021; Mayayise, 2024; Hussain et al., 2022).

This study investigated consumer behavior regarding online food purchase intentions on Amazon Egypt, focusing on the influence of eight factors: convenience motivation, service quality, price orientation, app quality, perceived risk, subjective norms, perceived control, and attitude. The hypothesis tests the significance of each factor in shaping consumer purchase intentions. Urbanization has led to reduced shopping time, prompting companies and retailers to adopt online shopping services to meet consumer needs.

Based on the theoretical foundations discussed, including both the positive and negative aspects of the relevant theories, a research model is developed to describe the conditions influencing consumer behavior in online food purchase intentions. The conceptual framework for this study is outlined as shown in Figure 1.

The survey measurements were derived from existing literature and were carefully adapted to suit the objectives of this research. The study examined and explained the relationships between the independent and dependent variables through hypothesis testing. Based on the literature review, a conceptual model was proposed. The questionnaire used in this study consisted of two main sections. The first section focused on quantitative data to investigate the effects of factors such as convenience motivation, service quality, price, app quality, perceived risk, subjective norms, and perceived control on consumer attitude and purchase intention. The measurement scale employed was the Likert scale, which is widely used to assess respondents’ level of agreement or disagreement with a series of statements. The Likert scale ranged from 1 (strongly disagree) to 5 (strongly agree). The second section of the questionnaire gathered demographic information, as shown in Table 1.

This study employed a quantitative design and was conducted in August 2024 in Egypt using online questionnaires via Google Forms. The sample size was 400, with data gathered from a purposive sampling technique, where participants were selected based on specific criteria (Malhotra et al., 2020). The questionnaire was distributed via popular social media platforms in Egypt, specifically Facebook and WhatsApp, which are widely used in Egypt. It targeted residents of Cairo and its surrounding districts. To ensure the accuracy of the sample’s geographic focus, the questionnaire included a demographic section requiring respondents to specify their location (e.g., city or district). Furthermore, the survey was promoted within local online groups and communities that are specifically associated with Cairo and neighboring areas. This approach helped to ensure that the responses were primarily from individuals residing in the intended geographic region.

Data analyses were performed using Statistical Package for Social Science (SPSS) version 25 and Analysis of Moment Structure (AMOS) Version 25 SPSS was used for descriptive analyses to analyze the characteristics of participants, and visualize the responses received. Then, AMOS was applied to test structural equation modeling (SEM) analysis through a two-stage procedure (Kline, 2023). In the first step, a confirmatory factor analysis (CFA) was conducted to evaluate the reliability and validity of the measurement model. In the second step, the full structural model was measured to evaluate the path analysis and the hypothesized relationships, with the help of standardized regression coefficients (β), t-values, and p-values, Smart PLS4 used for graphical path analysis.

Profile of the Respondents presents general information regarding the sample’s demographic features, The data in Table 1 indicated that the majority of respondents are young (52.8%), male (59.5%), single (69.2%), and educated, primarily earning less than 5,000 EGP per month, with most identifying as Muslim (91.5%) and holding a bachelor’s degree (63.3%), as shown in Table 1.

The reliability of key constructs in this study was assessed using Cronbach’s alpha and Composite Reliability (rho_c), with both measures indicating satisfactory internal consistency across all variables. Cronbach’s alpha values range from 0.754 to 0.857, suggesting that the scales used for the measurement of the constructs exhibit good internal consistency. Generally, values above 0.7 are considered acceptable for Cronbach’s alpha, with values exceeding 0.8 indicating strong reliability (Xu et al., 2024). These results gave a confirmation of the good reliability and validity of the research instruments as shown in Table 2.

The cross-loading analysis in the PLS-SEM model assesses the relationships between indicators and their corresponding constructs, as well as their associations with other constructs. In PLS-SEM, an indicator’s loading on its intended construct should ideally exceed 0.7, as this value indicates that the construct explains more than 50% of the variance in the indicator, demonstrating strong explanatory power and confirming the construct’s validity. On the other hand, cross-loadings on other constructs should generally remain below 0.5 to ensure discriminant validity and prevent ambiguity in the measurement model. This distinction is critical for maintaining clarity and robustness in the model’s interpretation. These criteria are well-supported by established guidelines in the literature (Afthanorhan et al., 2020), as illustrated in Table 3.

As shown in Table 3, the cross-loadings for each latent variable are higher than those for other latent factors, indicating that the latent variables exhibit discriminant validity.

Fornell-Larcker criterion—discriminant validity displays the criterion for discriminant validity, which is used to confirm that the constructs in the analysis are sufficiently distinct. According to this criterion, the square root of the average variance extracted (AVE) for each construct should be larger than the correlations between that construct and the others (Panzeri et al., 2024). In Table 4, the diagonal values correspond to the square roots of the AVE, while the off-diagonal values represent the correlations between the constructs. All diagonal values exceed the off-diagonal correlations, indicating that the constructs are distinct from each other. As a result, the measurement model satisfies the discriminant validity requirement, confirming that the constructs reflect separate underlying concepts. As illustrated in Table 4.

Testing the inner model or structural model aims to evaluate the relationship between latent constructs, significance values and R-square of the determined research model. Based on the research model that has been determined and has been tested by PLS-SEM analysis, the results of testing the inner model are as follows in Figure 2.

The Partial Least Squares Structural Equation Modeling (PLS-SEM) path model demonstrates that 59.8% of the variance in Attitude (R² = 0.598) is explained by the predictors PR, SN, APP, PC, CM, PSO, and SQ. Additionally, 77.6% of the variance in Purchase Intention (R² = 0.776) is accounted for by Attitude. The model incorporates latent variables (represented by blue circles) and their corresponding observed variables (depicted as yellow rectangles). All factor loadings exceed the threshold of 0.7, indicating strong measurement reliability (Afthanorhan et al., 2020). Path coefficients, represented by arrows, illustrate the directional relationships and effect sizes between independent and dependent variables, as depicted in Figure 2.

The value of each parameter seen from its significance value shows information related to the relationship between the variables used in this study. Testing the hypothesis in this study is to use or refer to the output path coefficients values as shown in Table 5.

Hypothesis testing can be done in two ways, testing using t-statistics and p-value. For t-statistic test, if the value of t ≥ (1.98) (André and Reinholtz, 2024), then the research hypothesis is accepted. Based on Table 5, 2 of the 8 constructs have a t-statistics value smaller than t-table, namely Service Quality to Attitude constructs of 0.308, and Perceived Risk to Attitude constructs of 0.136.

Testing the statistical relationship of variables can also be seen from the value of the coefficients path for each variable relationship showing a positive and significant relationship because the p is less than 0.0525 (André and Reinholtz, 2024). With this test it can also be seen that there are 2 constructs that are rejected among others; Service Quality to Attitude with p 0.756, and Perceived Risk to Attitude with p 0.896.

Based on the hypothesis test above, it can be concluded that the hypotheses H2, and H5 are rejected while the hypotheses H1, H3, H4, H6, and H7 are accepted.